JP2004535133A - Method for optimal use of SCTP in MPLS networks - Google Patents

Abstract

A method is proposed for exchanging information over a network between two entities. In this method, a first protocol is used to transmit information between networks, characterized by a label, wherein a target entity and a source entity are uniquely detected, and a first protocol is used. Use the underlying second transport protocol. The transport protocol guarantees receipt of the information by means of confirmation information. The method also includes a first initialization step in which labels 1 to n of the network path are exchanged via the first network path, and a second information exchange in which information is exchanged by the network paths 1 to n. Steps are used. Advantageously, the first protocol is MPLS, each FEC representing a unique target entity and a unique source entity. Advantageously, the second protocol is SCTP, wherein the IP address is replaced by the label of the network path.

Description

【Technical field】
[0001]
SCTP (Stream Control Transport Protocol, Stream Control Transmission Protocol) [RFC2960] is a transport protocol that provides high network redundancy. That is, it is guaranteed that all transmitted data arrives in the correct order. Instead of using only one possible transport path, such as TCP [RFC793] (implemented by IP [RFC7911]), multiple transport paths are used. This is necessary in applications where very high availability must be achieved, such as signaling in PSTN networks. At this time, IP is used as a network layer. SCTP supports IP with appropriate (generic) parameters in messages exchanged between two partners during connection establishment. In today's "IP" core networks, such as those used in wireless communications using UMTS, pure IP is increasingly omitted. Rather, direct LSPs (Label Switched Paths) using MPLS (Multi Protocol Label Switching) [RFC 3031] between individual network elements also use traffic engineering. Established. Known methods are published in the literature [RFC2960] and [RFC3031]. This is especially the case ... (where more information about traffic engineering is needed).
[0002]
It is an object of the present invention to use SCTP optimally in MPLS networks.
[0003]
This task is characterized by the features of the independent claims, in particular by the label, a first protocol used for the transmission of information between entity network paths uniquely detecting the target entity and the source entity. By transmitting information over a network between two entities. This first protocol is used only to route information to the target entity. The task of checking whether information is also actually arriving cannot be undertaken. For this, a second transport protocol is required. A second transport protocol based on the first protocol confirms receipt of the information by confirmation information sent back to the source entity. The method of the present invention does not prepare only one path for transmitting information, but prepares a plurality of paths. This must be transmitted when the communication is established.
[0004]
In a first initialization step, labels 1 to n of another network path are exchanged via the first network path. Each of these network paths allows a unique exchange of information between the entities. After the network paths for communication and information exchange are exchanged, the information to be transmitted is transmitted via network paths 1 to n. The information to be transmitted can be distributed to the network path with a smaller load according to the load situation or the response characteristic of each network path.
[0005]
To avoid collisions and increase throughput, network paths are used only for unidirectional information exchange. Bidirectional use is also conceivable, but does not correspond to SCTP settings.
[0006]
In a preferred embodiment, the first protocol is MPLS. The detailed structure of the protocol can be found in [RFC3031]. This protocol is further described below in principle. Basically, network protocols are represented by individual layers. The MPLS protocol is now located at layer 2. This protocol is usually provided with a layer 3 protocol in order to guarantee the uniqueness of the receiving and transmitting entities of the information packet. In this layer 3, TCP or alternatively SCTP is used as a transport protocol in layer 4. The two protocols guarantee that the information packet has actually arrived by transmitting a new packet when the acknowledgment of the receiver arrives at the sender in a predetermined time window. According to the invention, layer 3 and thus overhead can be omitted.
[0007]
In the MPLS network, so-called “transfer equivalent classes” (FECs) are used. This class is used to map a given IP address to a label. On the other hand, if there is a bijektiv relationship between the IP address and the equivalent class, the label can be used to carry out layer 3 tasks. That is, only one label for the sender IP address and the receiver IP address is used. Thus, each FEC represents a unique target entity and source entity. Normally, in SCTP, a number of IP addresses that are monitored as target addresses or source addresses and exchange information are exchanged.
[0008]
Thus, an entity has multiple IP addresses. Called a multihomed entity. Details regarding SCTP can be found in RFC 2960 or below. The entity is usually a server or a known computer. However, other network elements, such as routers, are also conceivable.
[0009]
The transport protocol of the method according to the invention is SCTP, where the IP address is replaced by the label of the network path. In another embodiment, a special path Id label is employed that is placed on the stack of information packets and uniquely maps information to individual source and target entities. Thus, unlike the alternative embodiment described above, the selected route does not need to be unique. The use of the path Id label can omit the IP address because the receiving side can identify where the information comes from based on the path IP label.
[0010]
Transport protocols basically guarantee the task of arriving information and keeping its order preserved. For this purpose, a known method of SCTP, which is also known from TCP, is used. Each packet containing the fragmented information has a number. The configuration order can be confirmed based on this number. A TSN (Transmission Sequence Number) is assigned to each information packet. Each response of the receiving entity or the target entity contains the TSN previously transmitted from the transmitting entity or the source entity. This makes it possible to confirm whether the packet has actually arrived or has been lost. If the response does not arrive within the RTT (round trip time), the packet is newly transmitted. RTO (Retransmission Timeout) defines a time interval, at the end of which a new transmission of the packet takes place.
[0011]
Before a network path for exchanging information can be used, it must be established. To this end, known methods for traffic engineering, which operate with IP addresses, are preferably used, and the router / switch is configured such that the target and source entities are unique for the path. The detailed function of this method is. . . (Not enough information).
[0012]
In another embodiment, label stacking is used during transmission. Each data packet is assigned a stack in which a label is monitored. Each label represents an equivalent class as described above. When a data packet with a label arrives at the router, the top label is removed from the stack and replaced with a new label corresponding to this equivalent class at the current router. Therefore, based on the start label and the end label, the path or route of the packet can be designated in advance with the information. SCTP selects different network paths when transmitting information depending on the throughput or response time. Therefore, a constant high throughput is guaranteed.
[0013]
Another component of the invention relates to an apparatus for exchanging information over a network, comprising means for implementing the method according to the invention. Advantageously, the device is a terminal in the form of a computer with a network interface. Furthermore, this device is a router that exists at the starting point of the route. This device must be able to assign labels to the corresponding terminal equipment.
[0014]
The drawings show possible embodiments of the invention. here,
FIG. 1 shows a network having a router that uses label stacking. Each router exchanges the label of the preceding router and replaces it with its own unique label, with the additional use of the path ID label located in the stack to ensure uniqueness.
FIG. 2 shows a network with a router that uses label stacking. Each router exchanges the label of the preceding router and replaces it with its own unique label, where the label represents a unique route.
FIG. 3 shows a network representing the prior art. The IP address is located in the data packet.
[0015]
FIG. 1 shows an MPLS network having two entities 11. These entities 11 are interconnected by two paths. The data packet 12 is directed via a router 16 through the network to arrive at the corresponding target entity 11. The route is characterized by a route Id label. The first route has a route Id label “route ID1”. The second route is characterized by a route Id label "route ID2". Each data packet 12 has one stack on which the label resides. An FFC is detected by the uppermost label 15. At the first transmission, the data packet was assigned a label with the number 0. The first router exchanges this label for label one. The last router in this path exchanges the label with the label having the number four. When responding, the second path is used. Here, too, corresponding labels are provided on the stack of data packets, and these labels are exchanged by the respective routers. The labels that determine the route are never exchanged, so the receiver can unambiguously know where the packet was sent from.
[0016]
FIG. 2 shows an embodiment without a path ID label. However, here, it is assumed that each label is associated with one unique route. A path is unique whenever the source and target entities are unique.
[0017]
FIG. 3 shows the prior art where the IP address is located in the information 14 to be transmitted. If, as usual, the label is not unique, but simply detects an equivalent class, it may be necessary to analyze the IP address in the information to be transmitted for the routing decision. This analysis can be very tedious and requires additional logic in the router.
[0018]
In the following, the individual protocols and their variants are considered in detail.
[0019]
In an MPLS network, a packet moves from one router to the next. Each router makes independent decisions on forwarding. That is, each router analyzes the header of the packet, and each router executes a program using a router algorithm. Each router selects a new route depending on the result of the router algorithm. Therefore, the selection of the next router is performed in two steps. The first step divides the total amount of possible packets into a number of equivalent classes (FEC). The second step maps each FEC to one route. Regarding the transfer decision, packets belonging to the same FEC are not distinguished. Different packets belonging to the same FEC cannot be distinguished. Packets with different target or source addresses are considered different packets. However, in order for MPLS to be usable in the present invention, the path and hence the equivalent class need to be unique. That is, the equivalent class represents a unique source entity and a target entity or their addresses.
[0020]
In the MPLS network, the association with the FEC is performed only once, that is, when a packet arrives at the network. The FEC associated with one packet is coded as a short value called a label. When the packet is transmitted to the next route, the label is transmitted together. Subsequent routers do not analyze the packet further. Only the label is inspected. The label is used as an index for the table from which the next route and next label can be obtained. The old label is replaced with the new label and the packet is forwarded to the next route.
[0021]
In MPLS networks, forwarding is controlled only by label. This has a number of advantages. In other words, the router only needs to have a slight performance. The router need only be able to analyze the label and check in a table which route is assigned to this label in order to replace the old label with the new label. Furthermore, high throughput can be realized by this simple task. Further benefits can be obtained from RFC 3031.
[0022]
The following defines some principles. Labels are short, local and characteristic identifiers and have a fixed length to identify FEC. The label is used to indicate the FEC to which the packet is associated. In basic use of the FEC, the FEC is assigned based on a network layer target address. However, the coding of network addresses is not a problem in the original use of FEC. In this respect, a difference from the present invention occurs. The coding of network addresses is problematic because labels are uniquely assigned to unique paths.
[0023]
To ensure that routers map packets to the same equivalent class, routers must regularly exchange information that knows which packets are assigned to labels. Furthermore, it is important that the same label is not used for different routers, in which case unambiguous identification of the preceding router is not possible. Furthermore, it should be mentioned that different processing is performed on the upstream and downstream. That is, these streams never have the same label. In the MPLS architecture, the decision to bind a given label to a given equivalent class is made by a router that is downstream with respect to this binding. The downstream router informs the upstream router about this connection. This information can be diverted to another packet as, for example, piggyback (Huckepack) information.
[0024]
In another embodiment, MPLS supports hierarchy, where the processing of labeled packets is completely independent of the hierarchy level. A packet without a label can be considered a label with an empty stack. The use of stacks becomes apparent when referring to packet tunnels. Such a tunnel can be obtained from document RFC 3031. When a packet is guided by a network path between two routers, the packet is always tunneled, where the network path also includes multiple routers.
[0025]
For example, if an explicit route including routers R1 to R4 is set in advance and a route including routers R1.1, R1.2, and R1.3 exists between routers R1 and R2, Another label is pushed onto the stack by router R1. Routers R1.1, R1.2, R1.3 operate on this new second element. As soon as the packet arrives at router R2, the top element is popped off the stack. This is a problem if the label is not on the stack. In a typical MPLS architecture, a network address (usually an IP address) is analyzed to detect an equivalent class. This situation must not occur when using the present invention. MPLS provides two types of route selection. One route selection establishes the route at the point of departure. Individual routes that must be traversed are detected. This is an explicit route. Since routers are not explicitly determined in the hop-by-hop route, each router can determine which router should be the next router based on the table. The invention can be operated with two route selection possibilities.
[0026]
SCTP is essentially a reliable transport protocol that constructs a packet network like IP. However, this IP is not used in the embodiment of the present invention. This can save layer 3 of the network schema.
[0027]
The advantages of SCTP are:
1. 1. Confirmed error-free transmission of user data without means Data fragment in the range of MTU (Maximum Transmission Unit). The MTU indicates the amount of packets that can be transmitted further in a packet-based network (eg, TCP / IP). TCP / IP uses the MTU to detect the amount of packets in each transmission. If an excessively large packet occurs (the router can no longer transmit), it is sent back to the corresponding computer.
3. Data transmission over multiple paths. These data have consecutive numbers to prevent transmission of information that does not match the original order.
4. High error resilience by supporting multi-homed entities at both ends of the association.
[0028]
In the original embodiment, SCTP was viewed as a layer between the user application and a connectionless packet network such as IP. However, in this embodiment, the IP network is redundant. This advantage has already been described above. SCTP, by its nature, is a connection-oriented association between two entities, but from its concept is designed more extensively than TCP.
[0029]
The association is initiated by an SCTP user query. A cookie mechanism similar to that described in RFC 2522 is used during association initialization to ensure protection against security attacks. If the user further desires to terminate the connection, the connection is terminated. Therefore, the half-open state as known from TCP is excluded. When the connection is closed, no new data will be received.
[0030]
The number of data streams in SCTP is confirmed at the time of connection initialization. The number is ascertained by two entities. The information exchanged via this stream has a corresponding number. Internally, SCTP assigns a unique stream sequence number to each information or each information packet. Using different streams ensures that information is transmitted even if one of the streams is blocked. In the present invention, a stream corresponds to an individual network path characterized by a label. A TSN (transmission sequence number) is assigned to each data packet. TSN does not depend on each stream sequence number. The receiving or target entity signals receipt of the packet by sending back a TSN. This ensures that all packets arrive. If the packet has disappeared, this packet is newly transmitted. Further, it is confirmed by the timer whether or not the packet should be newly transmitted. If the response time does not comply, the timer is set high and the packet is sent anew. The timer is set as high as possible until a response to the packet is obtained.
[0031]
The check field checks the packet for consistency.
[0032]
The following describes the establishment of a connection, which suggests that the known SCTP is different from the modified protocol according to the invention.
[0033]
Before an initial data transmission can take place between two SCTP entities (A and Z), a complete initialization process in which the SCTP association is formed must be performed.
[0034]
After the association is established, a unidirectional data stream can be transmitted over the path. The initialization process can consist of the following steps. Entity A first sends an initialization packet to entity Z. This initialization packet includes a check tag in a corresponding field (tag_A). The tag is a corresponding random number. After the initialization packet is transmitted, a timer is started. Entity A transitions to a cookie waiting state. Entity Z responds immediately with an initialization response packet based on this. Here, labels or route ID labels are exchanged instead of IP addresses. Route ID labels are exchanged whenever the route represented by the label is not unique. If the target entity cannot be unambiguously detected at the last router, the route is not always unique. This is always the case if the last label's equivalent class is not unique, ie it is a single shot. Entity Z needs to send the check field tag_A and tag_Z and the cookie as confirmation. Upon receiving this confirmation, the timer is stopped by entity A. Similarly, the state changes. Based on this, a cookie response is sent to entity Z notifying that the cookie has been obtained.
[0035]
Subsequently, the timer is newly started by the entity A. This timer is stopped upon confirming that Entity Z has received the cookie response. The interruption of the connection must always be notified to the corresponding entity. When a connection or association is so established, data is exchanged over the path. Inspection information is regularly transmitted to monitor the connection. As a result, the state of the connection can be detected.
[0036]
Further details can be obtained from RFC 2960.
[Brief description of the drawings]
[0037]
FIG. 1 is a network with a router that uses label stacking.
FIG. 2 is a network with a router using label stacking.
FIG. 3 is a network according to the prior art.

Claims (12)

In a method for transmitting information over a network between two entities,
Using a first protocol, characterized by a label, wherein the target entity and the source entity use a network path uniquely detected to transmit information between said entities;
Using a second transport protocol based on said first protocol, said transport protocol guaranteeing receipt of information by confirmation information;
Using a first initialization step of exchanging labels 1 to n of the network paths that allow information exchange between entities via the first network path;
A method for transmitting information over a network between two entities, characterized by using a second information exchange step of exchanging information over 1 to n network paths. The method of claim 1, wherein the first protocol is MPLS, and each FEC represents a unique target entity and a unique source entity. 3. The method according to claim 1, wherein the second transport protocol is SCTP, and wherein an IP address is replaced by a label of the network path. In a method for transmitting information over a network between two entities,
Using a first protocol, characterized by a label, wherein the target entity and the source entity use a network path for transmitting information between the entities uniquely detected by another path ID label,
Using a second transport protocol based on said first protocol, said transport protocol guaranteeing receipt of information by confirmation information;
Using a first initialization step of exchanging 1 to n path ID labels of network paths that allow information exchange between entities via the first network path;
A method for transmitting information over a network between two entities, characterized by using a second information exchange step of exchanging information over 1 to n network paths. The method of claim 4, wherein the first protocol is MPLS. The method according to one or more of the preceding claims, wherein the second transport protocol is SCTP and the IP address is replaced by a path ID label of a network path. 7. The method as claimed in claim 1, wherein a known method of SCTP, such as partial information number, TSN, RTO and / or RTT, is used to guarantee the transmission of the information. . Initialize the network path before establishing communication between the entities, advantageously applying known methods for traffic engineering operating with IP addresses, so that the target entity and the source entity are uniquely detected for one path The method according to one or more of the preceding claims, wherein the router / switch is configured as described. 9. Method according to one or more of the preceding claims, wherein different network paths are used when transmitting information depending on throughput and / or response time. The method according to one or more of the preceding claims, wherein label stacking is used in the MPLS protocol. The method according to one or more of the preceding claims, wherein the network path is used only for unidirectional information exchange. An apparatus for exchanging information via a network, characterized in that it comprises means for implementing the method according to one or more of the preceding claims.

Images